Analyzing and predicting the cost-effectiveness of Indian banks using hybrid network DEA and twin SVR approach
摘要
This research introduces a methodology for estimating the cost efficiency of decision-making units (DMUs) which incorporates directional distance function (DDF) with network data envelopment analysis (DEA) that can handle negative data and undesirable outputs. The proposed DDF-based network DEA approach considers the internal configuration of the DMUs with heterogeneous input costs, and exhibits unit and translation invariant properties. To enhance the predictive capabilities of the proposed methodology, twin support vector machine for regression (TWSVR) has been integrated to significantly reduce the computational resources required for complete re-execution of the complex network DEA model when new unit added, particularly in the context of large and growing datasets. Thus, offering valuable insights for strategic decision-making. The study extends its application to the Indian banking sector, encompassing three distinct sub-divisions to capture the inter-dependencies and interactions among components within a bank, providing a holistic evaluation of cost efficiency. The cost efficiency scores of the Indian banks operational in fiscal years 2011–12 to 2021–22 are estimated and predicted using hybrid network DEA window analysis (window width 5 years) and TWSVR approach. The findings conclude that on an average the banks outperform in transferring the available funds into loans and investments in comparison to generating funds. Moreover, the novel hybrid approach results in highly precise and accurate predictions of the system cost efficiency scores analyzed based on the evaluation metrics (mean square error: